A sample of music information retrieval approaches

Abstract

In this Perspectives edition of the Journal of the American Society for Information Science and Technology, we present articles specifically written as introductory overviews of ten important music information retrieval (MIR) research and development projects. As an MIR researcher myself, I am continuously awestruck by the multinational and multidisciplinary nature of MIR research. In this brief overview edition alone, we have authors from Australia, Canada, France, Great Britain, Italy, New Zealand, Poland, United States, Spain, and Taiwan. The disciplines represented both in this issue and in the broader MIR community include: general, music, and digital librarianship; computer science; audio engineering; signal processing; traditional information retrieval; musicology, music theory, and music psychology; the recording and music distribution business; human-computer interaction; and intellectual property law. Uniting this seemingly disparate aggregation is the common goal of providing the kind of robust access to the world's vast store of music-in all its varied forms (i.e., audio, symbolic, and metadata)-that we currently provide for textual materials.Yes, some MIR research teams do have visions of establishing the musical equivalent of Google.com. The enormous magnitude of the potential user base is too enticing not to be seduced by the monetary rewards associated with the successful launching of a Web-based MIR service. Beyond dreams of unlimited wealth, however, I believe most MIR research teams are motivated by two primary factors: (1) a basic love of music and (2) an overwhelming intellectual need to overcome the myriad difficulties posed by the inherent complexity of music. About the former factor, I have little to add but to say most, if not all, of the MIR researchers I have had the good fortune to meet have had long histories of amateur and professional music performance and/or scholarship. About the latter factor, I devote the following paragraph.Music information is a multifaceted amalgam that includes pitch, temporal (i.e., rhythm), harmonic, textual (i.e., lyrics, etc.), timbral (e.g., orchestration), editorial, and metadata elements. Music information is also extremely plastic. That is, any given work can have its specific pitches altered, its rhythms modified, its harmonies reset, its orchestration changed, its performances reinterpreted, and its performers arbitrarily chosen; yet, somehow, it remains the "same" piece of music as the "original." Because music information queries are founded in the same materials as music information, their creation and interpretation are also extremely plastic. Within this extraordinarily fluid environment, notions of "similarity" become particularly problematic. The problems associated with similarity in turn lead to difficulties in creating meaningful and useful outputs from MIR systems in response to the wide variety of potential music queries that will be submitted to them. Thus, the grand intellectual challenge facing past, present, and future MIR research ...

Information and higher things in life: Addressing the pleasurable and the profound in information science

“…as query input. In the extraordinarily fluid setting of music information retrieval, measuring "similarity" (between documents and the searcher's needs) becomes particularly problematic (Downie, 2004(Downie, , p. 1033.…”

“…In a feature issue of the Journal of the American Society for Information Science and Technology, Downie (2004) introduced the area of music information retrieval (MIR). He states:…”

Vagueness and uncertainty in information retrieval

“…A good deal of work on content-based image retrieval has been done [33]. Moreover, work has been done in the sound arena in terms of retrieval, including music [26].…”

Interactive information retrieval

“…Complex languages can be useful where the searcher wants to be very precise through the use of a detailed query (e.g., Pirkola, Puolamäki, & Järvelin, 2003) or where the data themselves are complex, for example, music data, which comprise different attributes such as timbre, tone, and pitch (Downie, 2004), each of which might be expressed as individual query components. Niemi, Junkkari, Järvelin, and Viita (2004) provide an example of the latter approach.…”